Epiq: Advancing smart audio for a fairer justice system
6 minute read
A custom virtual audio agent that joined hearings directly via native SDKs. We synchronised remote participants with courtroom microphones in real time, creating a single high-definition legal record. EpiQ is a global provider of technology-enabled legal services. As hybrid court proceedings became increasingly common, EpiQ engaged Insighture to enhance the quality and accuracy of remote testimony.
The Problem
As hybrid testimony became common, remote participants were being captured as compressed, single-track recordings, affecting transcription accuracy. There was also heavy reliance on expensive third-party tooling. In a courtroom environment, audio quality is directly tied to efficiency. Epiq faced a persistent technical bottleneck — remote participants on Zoom and Teams were being recorded as a single, compressed audio track. This made the transcription process unnecessarily extensive, often requiring a 4:1 ratio of manual work to decipher overlapping voices. EpiQ needed a way to turn these noisy, fragmented streams into a high-fidelity, 99% accurate record without relying on costly third-party licensing.
Outcomes Achieved
Key Services & Pillars
- Native SDK Integration
- Real-Time Speaker Identification
- Audio Synchronisation Engine
- Azure Cognitive Services
- Data Sovereignty Compliance
The Solution
Working as an extension of EpiQ's team, Insighture helped address the challenges associated with capturing and managing remote testimony in modern legal proceedings. Through a collaborative delivery approach, the teams developed a solution that improved the accuracy, reliability, and quality of digital court records. Native SDK Integration: Joining meetings via Microsoft Teams and Zoom Native SDKs to extract raw audio data directly. Multi-Channel Synchronisation: Merging remote streams with physical courtroom microphones into a single, high-definition audit trail. Technical Autonomy: Building a custom ecosystem from scratch to eliminate middleman costs and ensure compliance with data sovereignty requirements. Real-time speech-to-text integrates Azure Cognitive Services to produce a rolling transcript that updates as testimony is given. Speaker identification via a custom backend socket server tracks identities the moment a participant speaks, pushing their name directly to the UI.
“The team was a core part of our mission to deliver a fairer, more efficient system for legal recordings.”
— Manish Vazirani, CTO, EpiQ
The ability to have a single, high-fidelity view of the legal record is critical.
We were proud to engineer this and see it come together.